Your browser is not supported. Please update it.

29 May 2014

EBU damning LTE Broadcast study compares apples with oranges

The EBU (European Broadcasting Union) has called for further investigation of converged broadcast/cellular networks after dismissing current LTE Broadcast technology as too expensive and spectrally inefficient for mass distribution of linear TV services.

In an otherwise excellent study comparing costs and spectral efficiency between LTE Broadcast and second generation DVB-T2, carried out in collaboration with Germany’s public broadcasting research unit IRT (Institut f’r Rundfunktechnik), the EBU has been guilty of failing to take account of the multiple options mobile operators have for distributing linear video. In two separate but linked reports, one on spectral efficiency and the other on costs (both Opex and Capex), the EBU and IRT have focused merely on eMBMS (evolved multimedia broadcast multicast service), or MBSFN mode, used for LTE Broadcast in a direct head-to-head comparison with DVB-T2 with its commonly deployed HTHP (High Tower ‘ High Power) topology.

The reports ignore the other strings mobile operators and Telcos have to their bow, notably TD-LTE (Time Division LTE) and of course WiFi, to which a growing number are already offloading substantial amounts of unicast data. A key point arising at once here is that only the most popular channels would need to be delivered in broadcast mode, over mobile networks to prevent them falling over, while many others would be more effectively unicast, given that only one or two people are viewing them in any given cell, with many cells having no viewers.

In any case there will also be the potential for broadcasting over WiFi, with Cisco for one already enabling this in stadiums through its StadiumVision Mobile platform. This uses a form of multicast over WiFi to deliver live video to smartphones, tablets and PCs.

Meanwhile TD-LTE is coming up in the wings, having already been deployed quite widely in preference to the alternative FD-LTE (Frequency Division LTE) in the Asia Pacific region by China Mobile and Datang Telecom among others. In fact the global TD-LTE initiative was launched in 2011 by China Mobile in partnership with Bharti Airtel, SoftBank Mobile, Vodafone, Clearwire, Aero2 and E-Plus.

FD-LTE, which currently predominates in the rest of the world, especially Europe and the Americas, uses paired spectrum with one frequency band allocated for uploaded data and another for the download, which very much suits voice. TD-LTE by contrast allocates just a single frequency for all data, using time division multiplexing instead so that uploading and downloading alternates through time. This is more flexible in that the ratio between time allocated to uploading and downloading can be varied to take account of different application profiles, such as when there is a lot of unicast video streaming.

There are currently almost 30 commercial TD-LTE networks spanning 21 countries worldwide, with an additional 45 networks planned or in deployment, accounting for about 25% of the total. There is growing interest in Europe, where deployments are planned in both Denmark and Sweden. For this reason it is disingenuous of the EBU to ignore this factor, given the suitability of TD-LTE for combined multicast and unicast distribution of video traffic, with its ability to be fine-tuned accordingly.

The EBU has also largely ignored the broader business dimension, which is significant given the strenuous efforts being made by major Telcos to gain premium content rights that they can then exploit for a combination of fixed and mobile delivery. The key point here is that if LTE operators do succeed in gaining the content rights they want, they would be well placed to exploit these multiple delivery platforms to prove broadcasters wrong.

There is some worthwhile analysis within those two EBU/IRT papers that would reward study by both broadcasters and LTE operators. The analysis is based on simulations of LTE Broadcast in its usual Single Frequency Network mode (SFN), in which all cells reuse the same frequency bands for maximum spectral efficiency, using time synchronized wave fronts to avoid interference. By contrast standard LTE uses different bands for neighboring cells.

The simulations did have to take account of self-interference, the phenomenon whereby signals received by handsets are affected by the greater power of the transmitted signal. In two way unicast mode self-interference can be eliminated through sophisticated interference cancelling algorithms, but these require feedback from the handset and so do not work in one way broadcast mode.

Another point was that the standard propagation model was used for both LTE Broadcast and DVB-T2, to ensure that the comparison was totally fair. The significance of this is that usually for mobile networks, a less stringent criterion is applied, requiring a specified minimum percentage of the service area to be covered, allowing there to be some ‘dead’ spots with no signal. The standard propagation model gives a lower score for a given level of coverage, because it specifies a minimum probability of coverage for every location, which means that all locations must in effect get a signal some of the time. This might seem like hair splitting, but means that LTE Broadcast emerged from this study with a slightly lower score for spectral efficiency than it would from a traditional mobile analysis.

The analysis considered the principle reception use cases, fixed outdoor aerial, in-car, portable outdoor, near indoor reception (eg in a room with outdoor window), and deep indoor reception. The other variable was the ISD (Inter Site Distance), the gap between neighboring base stations, which ranged from 1 Km to 10 Km in the analysis. This yielded a curve relating ISD to spectral efficiency measured in bits per Herz of bandwidth for each of the five use cases. The result was around 3 bps/Hz for most scenarios at small ISDs of 1 to 2 Kms, but less for deep indoor, when strong degradation was found. The absolute limit of reception was an ISD of 5Kms for deep indoor, but for the other cases a baseline efficiency of 0.5 bps/Hz is maintained out to 8 to 10 Kms IBD.

However the EBU considers that a spectral efficiency of 4.5 bps/Hz should be the benchmark for fixed reception and 3 bps/Hz for portable reception. This would be achieved by a DVB-T2 network at an ISD of 5 Kms if it was deployed in an equivalent configuration to LTE MBSFN, according to the analysis. The upshot is that LTE Broadcast would only be capable of delivering robust linear services when deployed in the sort of dense networks only found in urban areas. In rural locations cellular ISDs are typically 5 to 10 Kms, when spectral efficiency of LTE MBSFN is between 0.2 and 1.5 bit/s/Hz, depending on the network topology and reception mode. This is not good enough.

The verdict appears even more damning for LTE Broadcast when it comes to the economics, considered in the second EBU/IRT paper. This showed that for very dense configurations LTE Broadcast would cost 25 to 30 times more to deploy than existing broadcast networks, although this comes down to 7 to 8 times for less dense configurations that would still deliver adequate spectral efficiency.

The cost ratio comes down further if deep indoor coverage is required, where LTE Broadcast scores more highly. For the dense 2 Km ISD, the analysis came out with annual opex of €3.1 billion for an LTE Broadcast deployment across the whole of Germany, coming down to €950 million for 5 Kms ISD. The whole study concluded on this basis that for now it was not economically feasible to provide mass linear TV content via cellular networks. It did not make sense on spectral efficiency grounds either because at the prevailing rural ISDs of 10Kms DVB-T2 did better. On the other hand at an ISD of 2 Kms LTE Broadcast scored a lot higher, as well as being better able to enable deep indoor reception.

But there was a caveat that rather changed the tenor of the argument, which was that the benefits of support for unicast transmission should be taken into account. For this reason the report concluded with the recommendation that the industry should ‘investigate further the possibility of reducing the eMBMS/MBSFN network costs by, for example, adding HTHP components to LTE networks, while maintaining or even improving the efficient use of the spectrum.’
We would go one further and say that a converged network capable of transmitting broadcast and unicast content is not just ‘a very attractive proposition’ as the report concluded, but is effectively what mobile operators are already starting to deploy with their heterogeneous network and carrier WiFi strategies. It is how the industry is moving and it could be broadcasters that are left behind.

There was no comparison offered on the costs of putting an extra chip into every phone to receive a separate signal (like the failed DVB-H standard required) compared to the fact that handsets already have the antenna and demodulation on-board in the baseband and app processor, to receive eMBMS signals. Which of course almost ruins the EBU argument entirely on its own. Neither was the fact that of course building a mobile network is immensely more expensive than building a broadcast network, but emulating an SFN across it is only a fraction of its use case, so the costs are effectively far lower. The plain truth is that mobile video represents an enormous threat to broadcast video, and the EBU is just beginning to see that threat in its rear view mirror, and it is pretty frightened.